Invariant Vorticity Decomposition (IVD)

• Invariant Vorticity Decomposition is a framework of methods that objectively isolates fluid rotation, shear, and stretching from the velocity gradient.
• It employs diverse techniques such as Liutex, normal–nilpotent splits, stochastic and covariant Helmholtz–Hodge decompositions to ensure invariance under different motions and transformations.
• The approach clarifies misconceptions by distinguishing true rotational motion from shear effects and by linking invariant measures to coherent structures and topological properties.

Invariant Vorticity Decomposition (IVD) appears in the cited literature as a family of distinct invariant or objective vorticity-based constructions rather than as a single standardized method. In different contexts it denotes Liutex-based rigid-rotation/shear splits, normal–nilpotent and real-Schur-form decompositions of the velocity gradient, convergent series expansions of the vorticity field, stochastic Lagrangian initial–boundary decompositions, covariant Helmholtz–Hodge splittings in acoustic geometry, APV-based wave–vortex projections, and topological classifications built from helicity and the Godbillon–Vey invariant. A separate terminological complication is that one paper states explicitly that, in its framework, “IVD” means “Instantaneous Vorticity Deviation,” not “Invariant Vorticity Decomposition” (Liu et al., 2021, Chen et al., 30 Aug 2025, Lam, 2013, Eyink et al., 2019, Park et al., 5 Feb 2026, Early et al., 2024, Machon, 2020, Haller et al., 2015).

1. Terminological scope and principal usages

The label “IVD” is attached to different mathematical objects depending on the problem class. In local kinematics, the emphasis is the decomposition of vorticity or of the velocity gradient into rigid rotation, shear, and stretching components. In analytical Navier–Stokes work, the emphasis is an invariant series decomposition of the vorticity field into space–time convolution constituents. In stochastic Lagrangian theory, the emphasis is a decomposition into backward-transported invariant contributions from initial data and from wall-generated vorticity. In compressible inhomogeneous media, the term is realized through a metric-covariant split of the metric-dual velocity one-form. In rotating stratified flow, invariant decomposition is tied to PV-inversion and energetically orthogonal wave–vortex constituents (Liu et al., 2021, Lam, 2013, Eyink et al., 2019, Park et al., 5 Feb 2026, Early et al., 2024).

Formulation	Core relation	Invariance statement
Principal/Liutex IVD	$$\boldsymbol{\omega}=\boldsymbol{\omega}_{\mathrm{rigid}}+\boldsymbol{\omega}_{\mathrm{shear}}=\mathbf{R}+(\boldsymbol{\omega}-\mathbf{R})$$	Galilean invariant (Liu et al., 2021)
NND-based IVD	$\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$	Coordinate-invariant; frame-indifferent under rigid motions (Chen et al., 30 Aug 2025)
Instantaneous Vorticity Deviation	$$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$	Objective under time-dependent rotations and translations (Haller et al., 2015)
Covariant CHHD/IVD	$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$	Metric-covariant (Park et al., 5 Feb 2026)
Stochastic Lagrangian IVD	$$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$	Backward martingale conservation law (Eyink et al., 2019)

This multiplicity of meanings is not incidental. Each construction isolates a different invariant structure: rigid-body rotation, rotational coherence, kernel-generated vortical constituents, stochastic Cauchy invariants, covariant vortical content, or PV-carrying balanced motion.

2. Local tensorial and kinematic decompositions

A major line of work defines IVD locally from the velocity gradient tensor $\mathbf{A}=\nabla\mathbf{u}$. In the principal decomposition in Cartesian coordinates, $\mathbf{A}$ is written as

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$

where $\mathbf{L}$ is the rigid-rotation (Liutex) tensor, $\mathbf{P}$ is the pure shear tensor, and $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$0 is the stretching/compression tensor. At the vector level, the associated invariant vorticity decomposition is

$\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$1

with $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$2 the Liutex vector. This construction is presented as Galilean invariant and is explicitly motivated by the statement that the antisymmetric or vorticity tensor cannot represent the fluid rotation or vortex, while the symmetric tensor cannot distinguish stretching and shear (Liu et al., 2021).

In the RS/Liutex framework, the velocity-gradient eigensystem supplies a real eigenvector $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$3 aligned with the local rotation axis when a complex-conjugate eigenpair exists. The Liutex magnitude is

$\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$4

Within this interpretation, $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$5 is twice the spatial mean angular velocity around the axis, $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$6 is the pseudo-time average angular velocity, and Liutex removes the non-rotational shear contamination from vorticity, leaving pure rotation quantified by $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$7 (Wang et al., 2018).

A more general kinematic theory reformulates IVD through the normal–nilpotent decomposition of the velocity gradient tensor. In the NND triad, the vorticity vector decomposes as

$\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$8

Here $\boldsymbol{\omega}=\mathbf{R}_N+\mathbf{s}_N$9 is the characteristic rigid-rotation mode and $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$0 is the characteristic spin/shear mode. The same work derives two extremal IVD modes and states that enforcing a minimization principle naturally yields the Liutex formula, with the Liutex vector identified as the extremal Case I mode $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$1 (Chen et al., 30 Aug 2025).

A related quadruple decomposition writes

$$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$2

separating dilation, axial stretch, in-plane distortion or normal rotation, and simple shear. In that framework, the mean rotation characterized by the vorticity $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$3 always consists of simple shear; the $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$4-criterion is sufficient but unnecessary for the existence of net rotation; the $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$5-criterion exactly identifies the existence of net rotation; and, when net rotation exists, $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$6 characterizes its absolute strength. The paper also gives the invariant-plane relation

$$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$7

which makes the role of simple shear in net rotation explicit (Li et al., 2024).

The real Schur form approach introduces yet another local split,

$$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$8

for which both decomposed vorticities are frozen-in to $$\mathrm{IVD}(\mathbf{x},t)=\|\boldsymbol{\omega}(\mathbf{x},t)-\overline{\boldsymbol{\omega}}(t)\|_2$$9 in the ideal setting. In the “uniform RSF in a fixed Cartesian coordinate frame” defined by $$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$0, $$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$1, and $$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$2, the model is described as in between two-dimensional-three-component and 3D3C flow, and the global helicity has only mutual linkage as the topological content, decomposed into two equal parts (Zhu, 2018).

3. Objective vorticity deviation and rotational coherence

In “Defining Coherent Vortices Objectively from the Vorticity,” the acronym IVD has a different meaning. The paper states explicitly that “IVD” stands for “Instantaneous Vorticity Deviation,” not “Invariant Vorticity Decomposition.” In three dimensions,

$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$3

and in two dimensions,

$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$4

Its Lagrangian counterpart is the Lagrangian-Averaged Vorticity Deviation,

$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$5

with the intrinsic dynamic rotation angle satisfying

$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$6

The paper proves objectivity under time-dependent rotations and translations by showing that the norm of the vorticity deviation from its spatial mean is frame-invariant (Haller et al., 2015).

This framework defines instantaneous Eulerian vortices by IVD level-set topology: vortex centers are local maxima of $$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$7, and vortex boundaries are outermost closed or tubular level surfaces of constant IVD encircling the center, with outward non-increasing IVD. Rotationally coherent Lagrangian vortices are defined analogously from LAVD. The same source emphasizes the Eulerian–Lagrangian link

$$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$8

along trajectories, so that rotationally coherent Eulerian vortices are “derivatives” in extraction time of LAVD vortices (Haller et al., 2015).

The framework is also used polemically against classical instantaneous criteria. Okubo–Weiss, $$\mathbf{u}^{\flat}=d\phi+\delta_\gamma\Psi+\mathbf{h}$$9, $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$0, and $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$1 are described there as not objective under time-dependent rotations and translations, whereas IVD and LAVD are objective by construction and directly measure equal instantaneous or net material rotation relative to the mean rotation of the fluid mass (Haller et al., 2015).

4. Series, stochastic, and translation-invariant decompositions

A distinct analytical use of IVD arises in the vorticity-based Navier–Stokes analysis on $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$2. Starting from a similarity-reduced Volterra–Fredholm equation, the vorticity is expanded into an infinite convergent series of “vortical elements,”

$$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$3

The same work states the integral invariance

$$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$4

and requires the decomposition to preserve aggregate invariant totals, including the spatial integral of the sum of vorticity components, derivatives of that quantity, helicity/Lamb-vector integrals, and impulse/angular impulse. The series counts $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$5, $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$6, and $$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$7 are given explicitly, and the construction is presented as a scale-dependent decomposition for transition and turbulence (Lam, 2013).

In wall-bounded stochastic Lagrangian theory, IVD is built from backward stochastic Cauchy invariants. For the stopped backward flow, the stochastic invariant

$$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$8

is a backward martingale, and

$$\boldsymbol{\omega}(x,t)=\mathbb{E}[\tilde{\boldsymbol{\omega}}_s(x,t)]$$9

At $\mathbf{A}=\nabla\mathbf{u}$0, the resulting decomposition separates contributions from initial data and from wall-generated vorticity selected by the hitting time $\mathbf{A}=\nabla\mathbf{u}$1. The same paper also gives a pointwise split into transport and stretching,

$\mathbf{A}=\nabla\mathbf{u}$2

and validates conservation of the mean Cauchy invariant for two selected buffer-layer events in turbulent channel flow (Eyink et al., 2019).

An algorithmic use of invariance appears in FMM-accelerated vortex methods based on the Lamb–Helmholtz decomposition

$\mathbf{A}=\nabla\mathbf{u}$3

The paper shows that the original representation is not invariant under translation and introduces conversion operators that restore it after shifting the expansion center. In that formulation, velocity and the vorticity evolution terms require evaluation of just two scalar harmonic FMM sums (Gumerov et al., 2012).

5. Covariant, wave–vortex, and state-space generalizations

In compressible, thermodynamically inhomogeneous media, IVD is realized by a covariant Helmholtz–Hodge decomposition with respect to an acoustic metric. The metric-dual velocity one-form is split as

$\mathbf{A}=\nabla\mathbf{u}$4

where the exact part is identified as the irrotational or acoustic component, the coexact part is the vortical or solenoidal component, and the harmonic component vanishes under the conditions used in the paper. The covariant vorticity is

$\mathbf{A}=\nabla\mathbf{u}$5

and $\mathbf{A}=\nabla\mathbf{u}$6 implies $\mathbf{A}=\nabla\mathbf{u}$7 for the acoustic component. For entropy-spot refraction and a normal shock, the paper reports that Euclidean post-processing produces significant leakage, whereas the covariant split remains at the numerical noise floor, typically $\mathbf{A}=\nabla\mathbf{u}$8, including near the sonic horizon (Park et al., 5 Feb 2026).

In rotating, non-hydrostatic stratified flow, invariant decomposition is tied to available potential vorticity and PV-inversion. The closed-form APV is written in terms of isopycnal deviation,

$\mathbf{A}=\nabla\mathbf{u}$9

and linearization gives the quasigeostrophic form

$\mathbf{A}$0

after discarding the nonlinear stretching term and the height nonlinearity. The paper then constructs energetically and enstrophically orthogonal constituents—geostrophic motions, internal-gravity waves, inertial oscillations, and a mean density anomaly—through projection operators that first perform PV-inversion and then project the residual onto wave modes (Early et al., 2024).

A broader nonequilibrium generalization appears for confined irreversible diffusions. There the relevant invariant objects are the invariant current $\mathbf{A}$1 and the stationary velocity $\mathbf{A}$2. The paper identifies two minimal descriptors, the $\mathbf{A}$3-weighted vorticity and the variation of $\mathbf{A}$4 along level sets of $\mathbf{A}$5, and proves the bound

$\mathbf{A}$6

In that setting, the decomposition is not a fluid-element split of $\mathbf{A}$7 into rigid and shear modes, but an invariant decomposition of steady irreversible circulation into vorticity-controlled and level-set-variation-controlled contributions to entropy production (De et al., 2023).

6. Topological and ideal-fluid invariant frameworks

In ideal incompressible Euler flow, an invariant-based IVD is built from helicity and the Godbillon–Vey invariant. Under the zero-helicity density condition

$\mathbf{A}$8

the one-form $\mathbf{A}$9 defines an integrable codimension-1 foliation to which the vorticity is tangent. Writing

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$0

the Godbillon–Vey invariant is

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$1

The paper states that, in this class, $\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$2 is a higher-order helicity-type invariant of the vorticity, becomes purely a vorticity invariant when nonzero, and measures helical compression and stretching of vortex lines. It also gives a vector formula

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$3

and derives both global and local obstructions to steady flow: $\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$4 is a global topological obstruction, while nonzero GV density implies $\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$5 locally (Machon, 2020).

Within this topological IVD, helicity defines a primary layer of classification, and the zero-helicity foliation class is refined by $\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$6, by the local conservation law for the GV density, and by distinguished closed vortex lines of Kupka type. The paper also gives an explicit formula

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$7

for a Thurston-style construction on $\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$8, showing that the invariant can be determined by the set of distinguished closed vortex lines and their slope data (Machon, 2020).

7. Comparison, invariance notions, and recurrent misconceptions

A recurrent misconception is that “IVD” has one fixed meaning. The supplied literature shows otherwise. In one line of work, “invariant” means Galilean invariance or coordinate independence of a local rigid-rotation/shear split; in another, it means objectivity under time-dependent rotations and translations; in another, it means preservation of integral invariants; in another, it means backward stochastic martingale conservation; in another, it means metric covariance with respect to an acoustic geometry; and in the topological setting it refers to invariance of helicity- and foliation-based quantities under volume-preserving diffeomorphisms (Liu et al., 2021, Haller et al., 2015, Lam, 2013, Eyink et al., 2019, Park et al., 5 Feb 2026, Machon, 2020).

A second misconception is to equate vorticity magnitude with rigid-body rotation. The Liutex and principal-decomposition literature uses simple shear as the canonical counterexample: for

$\mathbf{A}=\mathbf{L}+\mathbf{P}+\mathbf{D},$9

the vorticity is nonzero, but there is no local rigid rotation; in the Liutex framework, $\mathbf{L}$0, and in the principal decomposition the vorticity is purely shear (Wang et al., 2018, Liu et al., 2021).

A third misconception is to treat $\mathbf{L}$1, $\mathbf{L}$2, and $\mathbf{L}$3 as interchangeable vortex measures. The quadruple decomposition paper separates their roles sharply: $\mathbf{L}$4 reflects the strength of net rotation within the invariant plane relative to axial stretch and is only a sufficient condition; $\mathbf{L}$5 exactly identifies the existence of net rotation; and $\mathbf{L}$6 gives its absolute strength (Li et al., 2024).

A fourth misconception is the acronym clash itself. In the objectively defined vortex-coherence framework, IVD is not a decomposition at all, but the scalar field $\mathbf{L}$7 or $\mathbf{L}$8, whose level sets define instantaneous rotational coherence. That usage is mathematically precise and physically interpretable, but it belongs to a different conceptual family from Liutex-based, NND-based, stochastic, or topological decompositions (Haller et al., 2015).

Taken together, these formulations show that “Invariant Vorticity Decomposition” is best read contextually. In some settings it isolates rigid rotation from shear; in others it decomposes vorticity into invariant constituents across scales, trajectories, covariant geometry, or topological classes. The unifying feature is not a single formula, but the demand that the extracted structures be tied to a well-specified invariance principle.